PLS and NPAIRS Analysis The PLS (Partial Least Squares) and NPAIRS (Nonparametric, Prediction, Activation, Influence, Reproducibility, re-Sampling) neuroimaging software package developed at the Rotman Research Institute
Download

PLS and NPAIRS Analysis Ranking & Summary

PLS and NPAIRS Analysis Tags

analyze measure analyzer neuroimaging analysis npairs analysis measure task

PLS and NPAIRS Analysis Description

The PLS (Partial Least Squares) and NPAIRS (Nonparametric, Prediction, Activation, Influence, Reproducibility, re-Sampling) neuroimaging software package developed at the Rotman Research Institute PLS and NPAIRS Analysis is a set of tools that can be used in scientific purposes. Partial Least Squares (PLS), was first introduced to the neuroimaging community in 1996 (McIntosh et al., 1996), for measuring distributed task responses (Mean-Centering PLS and Non-Rotated Task PLS). PLS has also been applied to measuring distributed patterns that impact on task performance (Regular Behav PLS, Non-Rotated Behav PLS and Multiblock PLS) and finally to both task-dependent and resting state regional connectivity (McIntosh and Lobaugh, 2004).The NPAIRS (Nonparametric, Prediction, Activation, Influence, Reproducibility, re-Sampling) package was first introduced with canonical variates analysis (i.e., linear discriminant analysis) and a reproducibilty metric (Strother et al., 1997) followed by the addition of prediction metrics (Strother et al., 2002). NPAIRS uses a penalized PCA basis (PCA denoising) adapted to optimize the reproducibility and prediction metrics for CVA. In addition to measuring distributed task and resting state responses NPAIRS provides a statistical resampling framework with basic building blocks for benchmarking and comparing preprocessing and data analysis, (i.e., processing pipeline) choices (Strother et al., 2004).Both PLS and NPAIRS/CVA have proven to be robust methods for extracting distributed signal changes related to changing task demands in neuroimaging. Their relative strengths and weaknesses are currently being evaluated at the Rotman Research Institute. Requirements: · Java 1.6 or later What's New in This Release: · NPAIRS group partitioning bug fixed: Now it works even in the case that proportional group partitioning leads to fractional numbers of objects in a partition, e.g. if there are 2 groups of 5 to be partitioned into split halves (so a simple proportional split would lead to 2.5 members of each group in each split half), now one group is randomly chosen to be augmented (to 3) and the other one decremented (to 2) and vice versa for the other split half. · NPAIRS GUI improved to be compatible with more environments: PC Range 'Step' field should now be visible in Mac OS X and Ubuntu. · PLS behavioural bootstrap bug (off-by-one indexing error) fixed. · Option added to NPAIRS to use 'Run' instead of 'Session' as 'split object' (i.e. splitting unit) when resampling data. There is a new drop-down menu in the Analysis Setup window allowing user to specify split object (default is still 'Session'). This feature has been added to the GUI but is not yet included in command-line (batch) plsnpairs tools. · NPAIRS: Condition number of within-class CVA covariance matrix W is now always checked and warning is printed to analysis log file if cond(W) > 1000. (Condition no. is the ratio of highest to lowest eigenvalue in the spectral decomposition of W; the higher the condition no., the closer W is to singularity.) · NPAIRS CVA chi-squared computation convergence issue fixed: now if cdf calculation fails to converge for input values that are too high (and above critical value for default p-value threshold of 0.95) then corresponding chi-squared p-value is saved in output files as 1.0. · NPAIRS CVA R2 values saved in results .mat file as: npairs_result.r2_full_data (no. PC dims rows X no. CV dims cols) and MLCell npairs_result.r2_splits: one cell / CV dim and each cell containing (no. splits) rows X (no. PC dims) cols. Also saved as textfiles with suffix '.r2'. R2 values are calculated between CVA canonical scores for each CV dim and input data timeseries (e.g. PCA dims) to determine goodness-of-fit between each CV dimension and input (PCA) dims. · Blocked PLS results saved with suffix _BfMRIresult.mat to match Matlab PLS output syntax. Now results viewer will recognize which session files and datamats correspond to a given Blocked PLS analysis. · Mean-centred (Task) PLS permutation bug fixed: now result data giving Permuted (Singular) Values Greater than Observed should be correct.

PLS and NPAIRS Analysis Related Software